Rank One Computing Corporation (Nasdaq: ROC), which does business as ROC, said its latest facial age-estimation algorithm ranked first globally on three primary datasets in a National Institute of Standards and Technology evaluation. The result gives government technology teams a standardized performance reference as they assess age-assurance and identity tools, though it does not establish that the software is ready for any particular agency deployment.
ROC said its ROC-003 submission posted the lowest mean absolute error on the Visa, Application and Mugshots datasets for adults ages 18 to 30 in NIST’s Face Analysis Technology Evaluation for Age Estimation and Verification, or FATE AEV. Mean absolute error measures the average distance between an estimated and actual age; a lower result indicates a closer estimate. The company said it also retained a first-place position in NIST’s child-online-safety evaluation for ages 13 to 16.
The development matters because public agencies considering biometric or age-assurance technology need more than a vendor’s accuracy claim. NIST’s face-technology evaluations provide a common way to compare submitted algorithms across defined datasets and measures. That can give procurement and technical teams a starting point for asking whether a product’s reported performance matches the task, image quality and population involved in their planned use.
According to ROC, the algorithm produces an age estimate from a face image. The company did not describe its model architecture, training data, computing requirements, pricing, deployment model or interfaces with identity systems. It also did not say whether any government agency is using ROC-003 for age estimation. Those omissions matter because a ranking in a controlled evaluation does not answer operational questions about data handling, human review, record retention, accessibility or how a system behaves when an image is poor or a result is uncertain.
NIST describes FATE as an evaluation program for software that analyzes images, separate from its face-recognition work. The age-estimation track is intended to measure algorithmic performance rather than decide whether a technology should be used in a particular policy setting. Government buyers would still need to establish a lawful purpose, evaluate privacy and civil-rights requirements, test performance in their own operating conditions, and define when a person can challenge or override an automated result.
One potential public-sector application would be a state agency evaluating age-assurance options for an online service with statutory protections for minors. A standardized NIST result could help the agency narrow a field of prospective vendors before conducting its own security, privacy and usability testing. That example is illustrative; ROC did not identify such an agency deployment.
ROC’s result is more useful to public-sector buyers than a broad claim of AI accuracy because it is tied to a government-run benchmark and named datasets. It does not settle the larger procurement question. The company will need to show how ROC-003 performs within real agency workflows and governance controls, while agencies will need to decide whether an age estimate is appropriate evidence for the service at issue. For now, the result gives buyers a comparative data point—not a substitute for implementation testing or accountable policy.